The Messages are quite small (we send binary messages of about 150 Bytes via JMS/SwiftMQ).
We have a forwarding from one Router to another, second one is crashing, first one processing same messages does not crash (this one have 3 producers and one consumer).
We have about 20 connections on the router having the issue (using NIO), one producer (one connection) and one consumer. Therefore I do not think issue is related to consumer cache size (might be wrong ...), also there haven't been any changes (configuration/load/etc), also number of messages hasen't really changed (we talk about 350 transactional XA Msg/Second).
What is quite interessting is the usage of Survivor space, please see screenshot.
